Comparative Analysis on Solar Energy Forecasting Using Random Forest, XGboost, ARIMA, and Different Neural Networks

The rapid expansion of renewable energy sources, particularly solar power, has introduced critical challenges and opportunities in energy forecasting. Accurate predictions of electricity production from these sources are crucial for promoting sustainable energy use. In this study, we leverage a publicly available dataset from Kaggle, comprising historical records of solar electricity production, to assess the effectiveness of six distinct forecasting models: Convolutional Neural Networks (Conv1D), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Random Forest, XGboost, and ARIMA models. Our comparative analysis encompasses the training and evaluation of these models employing various performance metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R-squared). The results indicated that the Gated Recurrent Unit (GRU), a specialized recurrent neural network (RNN) architecture, demonstrated exceptional accuracy in predicting solar energy production, achieving the lowest Root Mean Square Error (RMSE) among the tested models with RMSE of 89.39. These findings highlight the importance of selecting model architectures tailored to the specific characteristics of energy production datasets, and they contribute valuable insights for enhancing renewable energy forecasting systems.